Search Console had 21 of 45 posts sitting in Discovered, currently not indexed. That status means Google knows a page exists and has chosen not to spend the time fetching it, which is usually a statement about how the site is put together rather than about the writing. The likely cause here was structural: there was no route to any article from the site navigation. Category archives could not serve as one, because they are deliberately set to noindex across the site. The new Insights page is the route, and it is now in the main menu.

• An index page earns its place through crawl distribution, not rankings. This page was never meant to rank, and designing it to would be a mistake. Its job is to sit one click from every page on the site and hand a crawler 137 links to articles that previously had no navigational parent. Those links are in the page source before any script runs, so what the crawler receives does not depend on JavaScript or on which topic filter a reader happened to select.
• Filtering had to be built so it cannot create pages of its own. The obvious way to build a topic filter puts the selection in the address as a query parameter, and that quietly manufactures a crawlable near duplicate of the page for every combination anyone clicks. On a site this size that competes with the very articles the page exists to promote. The filter state lives in the fragment instead, which no search engine treats as a separate address, and no entry is ever removed from the page when a filter is applied, only hidden.
• There was already content on the site that could not be reached. Ninety five question and answer pairs were written across the articles, but the article template showed only the first five of them while the structured data published all of them. Answers were therefore visible to Google and invisible to readers. All 112 pairs that exist today are now searchable and each one links to its own answer on the article that carries it.
• The measurement that matters has not happened yet. Shipping the page is not the same as clearing the backlog, and it would be easy to call this finished on the strength of the build alone. The test is whether that Discovered, currently not indexed count falls over the coming crawl cycles, and the reporting on it lags by around nine days, so it needs checking against live inspection rather than the summary. Until then this is a well founded change with its result still outstanding.